A Parts Based Registration Loss for Detecting Knee Joint Areas

TL;DR

This paper introduces a parts-based registration loss that improves the alignment of knee joint areas by encouraging spatial configuration similarity between reference and test images, aiding in medical image analysis.

Contribution

It proposes a novel parts-based loss function that automatically selects features and enhances registration accuracy for knee joint imaging.

Findings

Improved registration accuracy demonstrated on knee images

Automatic part selection from reference images

Enhanced spatial configuration matching

Abstract

In this paper, a parts based loss is considered for finetune registering knee joint areas. Here the parts are defined as abstract feature vectors with location and they are automatically selected from a reference image. For a test image the detected parts are encouraged to have a similar spatial configuration than the corresponding parts in the reference image.